|
Activity Number:
|
201
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #307345 |
|
Title:
|
Effects of Sample Survey Design on the Accuracy of Classification Tree Models in Ecology
|
|
Author(s):
|
Thomas C. Edwards*+ and D. Richard Cutler and Gretchen Moisen and Niklaus E. Zimmermann and Linda Geiser
|
|
Companies:
|
U.S. Geological Survey and Utah State University and U.S. Forest Service and Swiss Federal Research Institute WSL and U.S. Department of Agriculture
|
|
Address:
|
5230 Old Main Hill, Logan, UT, 84322-5230,
|
|
Keywords:
|
sample survey ; classification trees ; accuracy assessment ; probability samples ; non-probability sample
|
|
Abstract:
|
We evaluated the effects of probabilistic (DESIGN) and non-probabilistic (PURPOSIVE) sample surveys on classification trees models for predicting the presence of four lichen species in the Pacific Northwest, USA. Models derived from both survey forms were assessed using an independent data set (EVALUATION). Accuracies as gauged by resubstitution were similar for each lichen species irrespective of the underlying sample survey form. Cross-validation accuracies were lower than resubstitution for all species and both designs, and in all cases were closer to the true prediction accuracies based on the EVALUATION data. The classification tree structures also differed both among and within the species, depending on the sample survey form. We argue that greater emphasis should be placed on reporting cross-validation accuracies rather than resubstitution accuracies.
|
